Add a data source
Attach one or more data sources to a Data Collection so it has something to query. A collection can blend multiple sources behind one queryable surface.
Audience: builders
Open the Setup tab
After you create a Data Collection you land directly on the Setup tab, on the Sources step. The Setup workflow has three steps in order: Sources, Selection, Catalog. You cannot move forward until at least one source is attached.
Pick from sources IT has connected
Click Add data source
On the Sources step, click Add data source. It opens a popover anchored to the button (not a full dialog), so you can stay in context. The empty state shows the same action as a large tile next to Upload files.
Find the source
The popover lists every organization-managed connection your workspace can attach. Each row shows the connection name, the provider, and the current connection health. Use the search input to narrow by name.
Click Attach
Click Attach on a row to bind that source to this Data Collection. The popover stays open so you can attach more sources back-to-back. Click outside or press Escape to close it.
The supported providers come from IT's data sources catalog: Snowflake, Postgres, BigQuery, Databricks, Redshift, MySQL, ClickHouse, Microsoft SQL Server, Oracle, and TimescaleDB.
Data source versus Data Collection. IT connects a data source once for the workspace. You build a Data Collection on top of one or more of those sources. The collection is what Projects and Apps query.
When the source you need is missing
If the source you want is not in the popover, IT has not connected it yet. Two paths from here:
- Ask your workspace admin or team admin to connect it. Point them to Data sources for the provider list and connection flow.
- Confirm the connection landed in the active team. New connections are scoped to the team that connected them, so a connection set up in another team will not appear here.
Non-admins can still upload files. If you do not have permission to attach managed sources, the popover is disabled but Upload files stays available. Use the upload path for CSV, TSV, PSV, Excel, Parquet, Feather, or ORC files when you need to move quickly.
Blend multiple sources
A Data Collection is not limited to one source. Attach several when the questions you care about cross systems: GL transactions from the warehouse, vendor master from the ERP, contract terms from a file upload. The catalog reconciles the same thing across sources (vendor names, account codes, anything that shows up twice) when you build.
Keep the starting set tight. You can always add more later. The Sources step shows the count of attached sources and the count of connections still available to attach.
Detach a source
Each attached source has a Remove button on its card. Detaching takes the source out of the collection's selection and the next catalog build. It does not delete the source from the workspace, and it does not delete the data.
Detaching after a catalog build. If you remove a source that the current catalog references, your next build re-shapes the catalog without it. Active questions that depended on that source will start failing until you re-scope and rebuild.
What is next
With at least one source attached, click Continue to selection to choose the tables and columns this collection should include. See Scoping a collection.